The rapid evolution of e-commerce has exposed the limitations of traditional product retrieval systems in managing complex, multi-turn user interactions. Recent advances in multimodal generative retrieval -- particularly those leveraging multimodal large language models (MLLMs) as retrievers -- have shown promise. However, most existing methods are tailored to single-turn scenarios and struggle to model the evolving intent and iterative nature of multi-turn dialogues when applied naively. Concurrently, test-time scaling has emerged as a powerful paradigm for improving large language model (LLM) performance through iterative inference-time refinement. Yet, its effectiveness typically relies on two conditions: (1) a well-defined problem space (e.g., mathematical reasoning), and (2) the model's ability to self-correct -- conditions that are rarely met in conversational product search. In this setting, user queries are often ambiguous and evolving, and MLLMs alone have difficulty grounding responses in a fixed product corpus. Motivated by these challenges, we propose a novel framework that introduces test-time scaling into conversational multimodal product retrieval. Our approach builds on a generative retriever, further augmented with a test-time reranking (TTR) mechanism that improves retrieval accuracy and better aligns results with evolving user intent throughout the dialogue. Experiments across multiple benchmarks show consistent improvements, with average gains of 14.5 points in MRR and 10.6 points in nDCG@1.

AI-generated content technologies are widely used in content creation. However, current AIGC systems rely heavily on creators' inspiration, rarely generating truly user-personalized content. In real-world applications such as online advertising, a single product may have multiple selling points, with different users focusing on different features. This underscores the significant value of personalized, user-centric creative generation. Effective personalized content generation faces two main challenges: (1) accurately modeling user interests and integrating them into the content generation process while adhering to factual constraints, and (2) ensuring high efficiency and scalability to handle the massive user base in industrial scenarios. Additionally, the scarcity of personalized creative data in practice complicates model training, making data construction another key hurdle. We propose HLLM-Creator, a hierarchical LLM framework for efficient user interest modeling and personalized content generation. During inference, a combination of user clustering and a user-ad-matching-prediction based pruning strategy is employed to significantly enhance generation efficiency and reduce computational overhead, making the approach suitable for large-scale deployment. Moreover, we design a data construction pipeline based on chain-of-thought reasoning, which generates high-quality, user-specific creative titles and ensures factual consistency despite limited personalized data. This pipeline serves as a critical foundation for the effectiveness of our model. Extensive experiments on personalized title generation for Douyin Search Ads show the effectiveness of HLLM-Creator. Online A/B test shows a 0.476% increase on Adss, paving the way for more effective and efficient personalized generation in industrial scenarios. Codes for academic dataset are available at https://github.com/bytedance/HLLM.

Imagine decision-makers uploading data and, within minutes, receiving clear, actionable insights delivered straight to their fingertips. That is the promise of the AI Data Scientist, an autonomous Agent powered by large language models (LLMs) that closes the gap between evidence and action. Rather than simply writing code or responding to prompts, it reasons through questions, tests ideas, and delivers end-to-end insights at a pace far beyond traditional workflows. Guided by the scientific tenet of the hypothesis, this Agent uncovers explanatory patterns in data, evaluates their statistical significance, and uses them to inform predictive modeling. It then translates these results into recommendations that are both rigorous and accessible. At the core of the AI Data Scientist is a team of specialized LLM Subagents, each responsible for a distinct task such as data cleaning, statistical testing, validation, and plain-language communication. These Subagents write their own code, reason about causality, and identify when additional data is needed to support sound conclusions. Together, they achieve in minutes what might otherwise take days or weeks, enabling a new kind of interaction that makes deep data science both accessible and actionable.

With the increasing prevalence of multimodal content on social media, sentiment analysis faces significant challenges in effectively processing heterogeneous data and recognizing multi-label emotions. Existing methods often lack effective cross-modal fusion and external knowledge integration. We propose SentiMM, a novel multi-agent framework designed to systematically address these challenges. SentiMM processes text and visual inputs through specialized agents, fuses multimodal features, enriches context via knowledge retrieval, and aggregates results for final sentiment classification. We also introduce SentiMMD, a large-scale multimodal dataset with seven fine-grained sentiment categories. Extensive experiments demonstrate that SentiMM achieves superior performance compared to state-of-the-art baselines, validating the effectiveness of our structured approach.

Modern large language models (LLMs) have demonstrated impressive performance across a wide range of multi-step reasoning tasks. Recent work suggests that LLMs may perform planning - selecting a future target token in advance and generating intermediate tokens that lead towards it - rather than merely improvising one token at a time. However, existing studies assume fixed planning horizons and often focus on single prompts or narrow domains. To distinguish planning from improvisation across models and tasks, we present formal and causally grounded criteria for detecting planning and operationalize them as a semi-automated annotation pipeline. We apply this pipeline to both base and instruction-tuned Gemma-2-2B models on the MBPP code generation benchmark and a poem generation task where Claude 3.5 Haiku was previously shown to plan. Our findings show that planning is not universal: unlike Haiku, Gemma-2-2B solves the same poem generation task through improvisation, and on MBPP it switches between planning and improvisation across similar tasks and even successive token predictions. We further show that instruction tuning refines existing planning behaviors in the base model rather than creating them from scratch. Together, these studies provide a reproducible and scalable foundation for mechanistic studies of planning in LLMs.

We present a case study evaluating large language models (LLMs) with 128K-token context windows on a technical question answering (QA) task. Our benchmark is built on a user manual for an agricultural machine, available in English, French, and German. It simulates a cross-lingual information retrieval scenario where questions are posed in English against all three language versions of the manual. The evaluation focuses on realistic "needle-in-a-haystack" challenges and includes unanswerable questions to test for hallucinations. We compare nine long-context LLMs using direct prompting against three Retrieval-Augmented Generation (RAG) strategies (keyword, semantic, hybrid), with an LLM-as-a-judge for evaluation. Our findings for this specific manual show that Hybrid RAG consistently outperforms direct long-context prompting. Models like Gemini 2.5 Flash and the smaller Qwen 2.5 7B achieve high accuracy (over 85%) across all languages with RAG. This paper contributes a detailed analysis of LLM performance in a specialized industrial domain and an open framework for similar evaluations, highlighting practical trade-offs and challenges.

This paper investigates the capabilities of large language models (LLMs) in formulating and solving decision-making problems using mathematical programming. We first conduct a systematic review and meta-analysis of recent literature to assess how well LLMs understand, structure, and solve optimization problems across domains. The analysis is guided by critical review questions focusing on learning approaches, dataset designs, evaluation metrics, and prompting strategies. Our systematic evidence is complemented by targeted experiments designed to evaluate the performance of state-of-the-art LLMs in automatically generating optimization models for problems in computer networks. Using a newly constructed dataset, we apply three prompting strategies: Act-as-expert, chain-of-thought, and self-consistency, and evaluate the obtained outputs based on optimality gap, token-level F1 score, and compilation accuracy. Results show promising progress in LLMs' ability to parse natural language and represent symbolic formulations, but also reveal key limitations in accuracy, scalability, and interpretability. These empirical gaps motivate several future research directions, including structured datasets, domain-specific fine-tuning, hybrid neuro-symbolic approaches, modular multi-agent architectures, and dynamic retrieval via chain-of-RAGs. This paper contributes a structured roadmap for advancing LLM capabilities in mathematical programming.

Large language models have demonstrated remarkable versatility across a wide range of natural language processing tasks and domains. One such task is Named Entity Recognition (NER), which involves identifying and classifying proper names in text, such as people, organizations, locations, dates, and other specific entities. NER plays a crucial role in extracting information from unstructured textual data, enabling downstream applications such as information retrieval from unstructured text. Traditionally, NER is addressed using supervised machine learning approaches, which require large amounts of annotated training data. However, historical texts present a unique challenge, as the annotated datasets are often scarce or nonexistent, due to the high cost and expertise required for manual labeling. In addition, the variability and noise inherent in historical language, such as inconsistent spelling and archaic vocabulary, further complicate the development of reliable NER systems for these sources. In this study, we explore the feasibility of applying LLMs to NER in historical documents using zero-shot and few-shot prompting strategies, which require little to no task-specific training data. Our experiments, conducted on the HIPE-2022 (Identifying Historical People, Places and other Entities) dataset, show that LLMs can achieve reasonably strong performance on NER tasks in this setting. While their performance falls short of fully supervised models trained on domain-specific annotations, the results are nevertheless promising. These findings suggest that LLMs offer a viable and efficient alternative for information extraction in low-resource or historically significant corpora, where traditional supervised methods are infeasible.

Controlling high-dimensional and nonlinear musculoskeletal models of the human body is a foundational scientific challenge. Recent machine learning breakthroughs have heralded policies that master individual skills like reaching, object manipulation and locomotion in musculoskeletal systems with many degrees of freedom. However, these agents are merely "specialists", achieving high performance for a single skill. In this work, we develop Arnold, a generalist policy that masters multiple tasks and embodiments. Arnold combines behavior cloning and fine-tuning with PPO to achieve expert or super-expert performance in 14 challenging control tasks from dexterous object manipulation to locomotion. A key innovation is Arnold's sensorimotor vocabulary, a compositional representation of the semantics of heterogeneous sensory modalities, objectives, and actuators. Arnold leverages this vocabulary via a transformer architecture to deal with the variable observation and action spaces of each task. This framework supports efficient multi-task, multi-embodiment learning and facilitates rapid adaptation to novel tasks. Finally, we analyze Arnold to provide insights into biological motor control, corroborating recent findings on the limited transferability of muscle synergies across tasks.

User queries in real-world recommendation systems often combine structured constraints (e.g., category, attributes) with unstructured preferences (e.g., product descriptions or reviews). We introduce HyST (Hybrid retrieval over Semi-structured Tabular data), a hybrid retrieval framework that combines LLM-powered structured filtering with semantic embedding search to support complex information needs over semi-structured tabular data. HyST extracts attribute-level constraints from natural language using large language models (LLMs) and applies them as metadata filters, while processing the remaining unstructured query components via embedding-based retrieval. Experiments on a semi-structured benchmark show that HyST consistently outperforms tradtional baselines, highlighting the importance of structured filtering in improving retrieval precision, offering a scalable and accurate solution for real-world user queries.